Autonomous Driving (AD) has advanced significantly in recent years, yet widespread deployment remains limited. One of the most enduring challenges in Autonomous Vehicle (AV) development is planning a safe, comfortable, and efficient motion in complex, real-world environments. This thesis addresses motion planning across three distinct domains: urban shuttles, passenger vehicles, and truck-trailer systems. It contributes practical insights and novel approaches toward scalable autonomous mobility. The first part of this work presents an integrated motion planning framework for the Continental Urban Mobility Experience (CUbE) driverless shuttle. Extensive real-world testing over several years highlights the system’s robustness and underscores the importance of long-term validation in urban settings. Key innovations include a multi-layered planning stack and a data-driven motion forecasting approach that enhances interaction with human traffic participants. The second part investigates the behavior of human drivers in understructured traffic environments. Those are areas that fall between well-defined road systems and fully unstructured spaces. A novel analysis framework is introduced for mining patterns from naturalistic trajectory datasets, enabling AVs to better blend into human traffic and navigate ambiguous scenarios with improved predictability and safety. The final part of the thesis explores Deep Reinforcement Learning (DRL) for planning and controlling complex truck-trailer maneuvers. A DRL-based approach is developed and evaluated in simulated environments, demonstrating the method’s potential to handle the nonlinear dynamics of articulated vehicles. These contributions advance the field of motion planning by combining theoretical insights, system-level integration, and empirical evaluation. They offer pathways for improving AV behavior across diverse platforms and use cases, ultimately supporting the broader adoption of AD technologies.
